Metraipolis › archive › dc_bfd8db74fc1d

Towards Subject Consistency over Dynamic Subject Sets in Video Generation — arXiv:2610.01052

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.01052
License (as archived)
CC0-1.0
Archived
2026-10-06T19:12:20.099Z
SHA-256 of the archived text
9c91d8a87813115771257db28c3d41afa8ee1889462f10be928d4f248ac6784e
Stable URL of this page
https://metraipolis.com/documento/dc_bfd8db74fc1d

How to cite

Towards Subject Consistency over Dynamic Subject Sets in Video Generation — arXiv:2610.01052 — arXiv (en), https://arxiv.org/abs/2610.01052 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_bfd8db74fc1d (Metraipolis archive dc_bfd8db74fc1d, sha256 9c91d8a87813115771257db28c3d41afa8ee1889462f10be928d4f248ac6784e)
BibTeX
@misc{dcbfd8db74fc1d,
  title = {Towards Subject Consistency over Dynamic Subject Sets in Video Generation — arXiv:2610.01052},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.01052},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_bfd8db74fc1d; sha256 9c91d8a87813115771257db28c3d41afa8ee1889462f10be928d4f248ac6784e},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.01052. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

We argue that as video generation extends to longer durations, subject consistency should be evaluated over \textit{dynamic subject sets}. We therefore introduce \textbf{DynSC-Eval}, an evaluation framework that dynamically tracks eligible subjects throughout their visible lifespans and measures local continuity and global identity preservation using six complementary object-level metrics, with explicit detection of inconsistency events. To validate its effectiveness, we design synthetic experiments that actively inject inconsistency events, demonstrating both the sensitivity of DynSC-Eval and the limitations of existing metrics. Evaluations of diverse models on 5s, 15s, and 60s video generation further reveal substantial subject consistency differences that are obscured by conventional metrics. Beyond evaluation, we construct rewards from DynSC-Eval and apply DiffusionNFT post-training in an autonomous-driving testbed. On 5s generation, our approach reduces the six inconsistency metrics by an average of 13.82\% for Wan-2.1-1.3B and 5.66\% for SANA-2B, with improvements also observed on the I2V model ReSim. Qualitative comparisons further demonstrate the effectiveness of our method. We then extend generation to 10s and 30s through curriculum learning and show that consistency optimization remains effective while largely preserving other capabilities.

Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-10-02 in cs.CV).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2610.01052 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-06T17:37:36.729Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-10-02?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).